CSTA (Lisette Antigua): What is Responsible AI & The CARE Framework
What is Responsible AI and What Does That Look Like in the Classroom?
Primary Source: CSTA Article Page
Overview
Published by the Computer Science Teachers Association (CSTA), this article by CSTA Responsible AI Fellow Lisette Antigua presents a practical definition of Responsible AI for K–12 education and introduces the CARE Framework—a student-centered model designed to foster critical thinking, ethics, and human accountability when using AI tools across content areas.
Antigua highlights that banning AI is an ineffective strategy because students use these tools outside school regardless. Instead, educators must set explicit usage boundaries, reframe assessments around student thinking and reflection, and educate students on AI’s mechanics and limitations.
Core Definition of Responsible AI in Education
Responsible AI is the process of integrating AI in the classroom to enhance—not replace—learning, while maintaining accountability, transparency, and human oversight.
Key Pedagogical Principles
1. Mindset Shift & Explicit Usage Boundaries
Cheating and misuse happen when expectations are vague or when grades are prioritized over growth. Antigua sets clear classroom boundaries: * Allowed: Brainstorming ideas, debugging code, and conceptual clarification. * Prohibited: Summative exams and initial skill acquisition. * Required Disclosure: Students must transparently explain how AI helped, what outputs were accepted or rejected, and what modifications they made independently.
2. Process-Oriented Assessment & Reflection
AI can quickly generate boilerplate text or code, but it cannot explain personal reasoning, justify design decisions, or reflect on individual growth. By requiring students to show their thinking and iterative process, shortcuts lose their value.
3. Demystifying AI as a Cheating Deterrent
Teaching students how AI works—explaining that machine learning models are statistical pattern matchers that are neither “smart” nor inherently neutral or credible—is the strongest deterrent to blind reliance and academic dishonesty.
The CARE Framework
The CARE Framework provides a practical, 4-step process for critical evaluation of AI tools and dataset design:
+-------------------------------------------------------------+
| CARE FRAMEWORK |
+-------------------------------------------------------------+
| C - Check the Data (Sources, missing data, features) |
| A - Ask About Impact (Who benefits/harmed, stereotypes) |
| R - Review and Revise (Ethical iteration over pure accuracy) |
| E - Evaluate and Explain (Model cards, documentation) |
+-------------------------------------------------------------+- Check the Data
- Focus: Data sources, missing data, and training features.
- Classroom Tool: Code.org’s AI Lab allows students to select labels and features to observe how data shapes predictions.
- Ask About Impact
- Focus: Human impact, potential harm, stereotypes, and over/underrepresented groups.
- Principle: AI is never neutral. Fairness is determined by human choices, not automated algorithms.
- Review and Revise
- Focus: Iterative refinement, removing harmful features, shifting to behavior-based traits, and rebalancing datasets.
- Principle: Challenges the idea that higher accuracy alone equals better AI; ethical design prioritizes thoughtful choices.
- Evaluate and Explain
- Focus: Testing performance and equity, documenting trade-offs, and creating transparent documentation (mirrors industry model cards).
- Outcome: Transforms students from passive users into responsible technology designers.